Vast amounts of digital information are available to users through the interconnection of computers and storage by the Internet. Indeed, traditional paper records can be scanned and retained electronically. Emails or instant messages between a company's engineers, planners, or financial officers can have the same or greater significance as formal memorandums or journal articles in a knowledge database. Although available in great quantity, digital information must be accessed in order to be used efficiently and profitably. Typical search methods use queries consisting of a Boolean combination of words and phrases which return documents containing words or phrases that match the search query. Imprecision and ambiguities in the words and phrases affect the effectiveness of searches. Refining a search takes time and skill and a user's search capability usually improves with practice and experience.
However, businesses need fast and reliable search capabilities that are not dependent on a user's skill and experience. Speed and reliability are improved by a search capability that precludes a null result. A null result occurs when desired information is not present in the data base, or the desired information is present, but the search term fails to locate the information. Speed and reliability are improved by a search that is controlled so that users are logically directed to the sought after information without distractions by non-relevant information. Thus the record count of documents responsive to a search should never be zero, and should include only documents that are relevant to the search.
In addition to speed and reliability, businesses need a search capability that can restrict accesses to its total amount of digital information. For example, not all employees have a need to access all information. Trade Secret information, while part of the total universe of a business's digital information, must have access limited only to those with proper authorization. Moreover, one division of a corporation may not need access to another division's information. Employee files including health records and financial data must be protected in accordance with various State and Federal laws. Thus the need for information access with speed and reliability is affected by equally important needs to restrict access where appropriate.
Personnel responsible for organizing the universe of a businesses' digital information have found that a controlled search can be achieved by a faceted search using a facet tree. A facet tree is a categorization scheme for hierarchially structured data and content records where the facets and facet elements of the tree guide the search, and a posting list of records is associated with each facet element. A facet is a node in the hierarchial network and a facet may have any number of facet elements. The total number of records in a facet is the sum total of all records in the posting lists of the facet elements beneath the facet in the facet tree.
A faceted search is conducted by a faceted search engine. Selection of a facet (a word or phrase) causes the faceted search engine to present the next set of facets, and so on along the facet tree until a final facet is reached and the facet elements are displayed. When the user selects one of the facet elements, the documents on the associated posting list are displayed. The relationship of the facets to each other and of the facets to the facet elements is created by tagging. In general, a tag is one or more characters attached to a set of data that contain information about the set, including its identification. By tagging is meant assigning a value to a facet so that it is identified with another facet in a hierarchical relationship and assigning a value to a facet element so that it is identified with a particular facet.
Current approaches to faceted search are based upon the tagging of data and facet elements in a strict parent/child relationship. Using a faceted search engine, end users choose one facet or facet element at a time. Document or data record counts are displayed in conjunction with the facet elements, such that end users are aware of the total number of records and also how many data records exist at a given facet element within a facet. The total number of records means the number of records for all of the combined facets and facet elements that have been selected.
Faceted search has the advantage of eliminating null results when end users are searching for records across many combined dimensions. To do this, data records that are tagged to a facet element are also made available at facet levels. By way of example, a user might use a faceted search engine to locate content in geographical, industry, and technology facets by choosing “Europe,” “Retail,” and “Supply Chain” as facets. When one does this, all of the records in posting lists identified with the facets of Europe, Retail, and Supply Chain are aggregated and returned. The total number of records returned is the set intersection of all of the data records rolled up to Europe through Retail and Supply Chain.
Another form of faceted search is a conversational search. In a conversational search, end users are presented with a series of questions derived from or tagged to a facet hierarchy, and the answers given by the end users result in faceted responses rendered by a conversational search engine. The response may include search engine results at the record level finding, for example, documents, titles, abstracts, and other information about those documents. The response may also include navigation from the facet that the question was asked about, to a succeeding facet or facet element. In a conversational scheme, the semantic content of the questions matters, and so the order in which the questions are presented is important.
Current methods for tagging of content and data records for faceted and conversational search are tedious, manually intensive, and prone to error. When content is tagged by content providers, authors, or other editorial roles, it involves human selection of elements from a faceted categorization scheme. The manual tagging solutions currently offered by content management providers require selection of single elements in combination. This results in a much higher number of mouse clicks, and a greater amount of time for tagging each record or document to be presented via a faceted or conversational search rendering engine. Tagging documents manually is often cost-prohibitive because of the number of steps required to select each element in a number of deep hierarchical structures.
Businesses have a need to apply policies to tagging to meet requirements for restricted access for security and legal reasons, but also to apply policies to tagging to promote the efficiency of providing results to end users. Automated tagging algorithms exist such as the Eureka algorithm and software from IBM Research. Other automated tagging algorithms are known. For example, United States Patent Application 2002/0091661, entitled “Method and Apparatus for Automatic Construction of Faceted Terminological Feedback for Document Retrieval” (the '661 application), discloses a method for extracting key informational concepts or “facets” from a collection of documents. In the '661 application, facets are chosen heuristically based on lexical dispersion. As used in the '661 application, lexical dispersion is “a measure of the number of different words with which a particular word co-occurs within such syntactic contexts.” The facets and their corresponding lexical constructs provide a structured summary of the contents of a result set.
However, these automated solutions rely on factors such as word frequency in a document and cannot tag documents in accordance with business rules. In order to tag data records and documents in accordance with a stated policy, a person must manually perform the tagging according to specific business rules. While algorithms are capable of approximating human tagging based on semantic categorization, they cannot perform tagging in accordance with policies. Additionally, each time a tagging policy changed, the algorithm would need to be changed. Therefore, although labor intensive, human tagging is necessary to make the policy distinctions that cannot be made by a machine.
What is needed is a solution that allows for rapid selection of tagging facets and facet elements in a facet tree by eliminating repetitive user actions so that a user can quickly tag data records, documents, or collections of documents to a given facet tree via navigation of the elements in a tree format for a given facet.